- dovecot - dovecot.org

dsync between 2 instances on same machine (but different versions and different DNS hostnames) fails with "doveadm server sent invalid handshake"
by rmfuhrer＠optonline.net 26 Oct '25

26 Oct '25

I'm trying to use dsync to upgrade from an existing 2.3 server running native on MacOS to a 2.4.1 server running inside a Docker container on the same machine, using the official Dovecot 2.4.1 Docker image from DockerHub. For testing, I decided to set up a new hostname (CNAME) record in DNS pointing to the same IP, since both Dovecot instances are running on the same host. The new 2.4.1 Docker-containerized instance is working fine so far, which is to say I can connect to it from a mail client (the Mac Mail app) using TLS/SSL, and I'm able to move mail messages in, delete them, create folders, etc. For the next phase of testing, I want to migrate a single folder of a single user's email from the 2.3 instance to the new 2.4.1 Docker instance. I recently found out about dsync and thought I'd give it a try. The test Dockerized instance exposes its IMAPS port to a different port than the legacy instance (of course). I found that dsync seems to behave differently than the IMAP service wrt the certficate chain: Whereas the Mail client was able to negotiate the encrypted connection with the LetsEncrypt server cert alone in the dovecot cert file, with dsync, the 2.3 server seemed to require the entire cert chain to reside in the file. (?) After resolving that, the source dsync could at least get past the certificate validation phase, but then complained thusly: Oct 19 21:52:11 doveadm(myuser(a)mydomain.net): Error: doveadm server sent invalid handshake: * OK [CAPABILITY IMAP4rev1 LOGIN-REFERRALS ID ENABLE IDLE SASL-IR LITERAL+ AUTH=PLAIN AUTH=LOGIN] Dovecot ready. Oct 19 21:52:11 doveadm(myuser(a)mydomain.net): Error: Disconnected from remote: EOF Oct 19 21:52:11 doveadm(myuser(a)mydomain.net): Debug: auth-master: conn unix:/usr/local/var/run/dovecot/auth-userdb (uid=0): Disconnected: Connection closed (fd=8) I can't tell from that error message what the source dsync doesn't like about the destination dsync, or for that matter, whether it might be the destination dsync that doesn't like what the source dsync is saying?? Anyway, I'm invoking dsync like this, to sync just a single folder: $ sudo doveadm -D -v backup -u myuser(a)mydomain.net -m FolderName tcps:testserver.mydomain.net:31993 ALSO: I confess that I also don't understand how dsync can work without asking for user account passwords, if it's indeed running over the IMAPS service. How can the destination service trust the incoming dsync connection? I assume it's running over a normal IMAPS connection, since that's the only port that I've given to the source dsync command, but maybe that's not the case? The docs are pretty unclear about this aspect, but I'm no expert, so maybe this is just obvious to others. (?)

3 11

mdbox: double forward-slash in mail_path
by Edmund Lodewijks 23 Oct '25

23 Oct '25

Hello List! I am testing mdbox on Dovecot 2.4.1 (currently using Maildir, but thinking of converting). I created a test server with a test domain. Installed relevant software from default repos on Debian 13: - Postfix - Dovecot 2.4.1 - PostfixAdmin - MariaDB After I had created an email address, I checked the settings as follows: ``` # doveadm user mail@$REDACTED field value uid 5000 gid 5000 home /vhosts/$REDACTED/mail/ mail_path /vhosts/$REDACTED/mail//mdbox quota_rule *:bytes=1048576000 ``` As you can see, there is a double forward-slash in the mail_path line. From /etc/dovecot/local.d/10-mail.conf: ``` mail_driver = mdbox mail_path = ~/mdbox ``` This is based on the official Dovecot docs page (see at the very bottom): https://doc.dovecot.org/2.4.1/core/config/mailbox_formats/dbox.html Is this harmful, or just "cosmetic"? Have I made a mistake somewhere? I presume the mail_path line is not even required, as it is the default for that mail_driver. Thank you for your time and reply. Kind regards, Edmund Lodewijks -- Edmund Lodewijks <edmund(a)proteamail.com> TZ: UCT+2 / GMT+2

4 8

Connecting to multiple ldaps AD from Dovecot 2.4.1
by Grepl Zbynek 23 Oct '25

23 Oct '25

Hello List! I am testing auth-ldap.conf.ext on Dovecot 2.4.1 Debian 13 (trixie) Postfix 3.10.4 Dovecot 2.4.1-4 (7d8c0e5759) Everything works fine when I use only one ldap AD for authentication. I need to connect multiple ADs of our organizations which have separate ADs. The problem is that it always takes the last ldap_uris as working even if I put them in separate files. *configuration - auth-ldap.conf.ext* *### AD - ORG_ONE* ldap_uris = ldap://XX.YY.ZZ.CZ ldap_auth_dn = CN=ReaderAD,CN=Users,DC=domain1,DC=local ldap_auth_dn_password = secret_password_1 ldap_base = OU=ALFA,dc=domain1,dc=local ldap_version = 3 *### domena subdomain1.domain1.cz* passdb ldap { ldap_filter = (&(objectClass=user)(mail=%{user|username}@%{user|domain})) ldap_bind = yes result_internalfail = return-fail } userdb ldap { driver = ldap ldap_filter = (&(objectClass=user)(mail=%{user|username}@%{user|domain})) fields { mail_path=/home/vmail/%{user|domain}/%{ldap:mail}/Maildir mail_inbox_path=/home/vmail/%{user|domain}/%{ldap:mail}/Maildir/ } } *### AD - ORG_SECOND * ldap_uris = ldap://AXX.AYY.AZZ.ACZ ldap://BXX.BYY.BZZ.BCZ ldap_auth_dn = CN=ReaderAD,OU=ServiceAccount,OU=DELTA,OU=GAMA,OU=ALFA,OU=BETA,DC=domain2,DC=cz ldap_auth_dn_password = secret_password_2 ldap_base = OU=BETA,dc=domain2,dc=cz ldap_version = 3 *### domena subdomain2.domain1.cz* passdb ldap { ldap_filter = (&(objectClass=user)(mail=%{user|username}@%{user|domain})) ldap_bind = yes result_internalfail = return-fail } userdb ldap { driver = ldap ldap_filter = (&(objectClass=user)(mail=%{user|username}@%{user|domain})) fields { mail_path=/home/vmail/%{user|domain}/%{ldap:mail}/Maildir mail_inbox_path=/home/vmail/%{user|domain}/%{ldap:mail}/Maildir/ } } *--- doveadm auth test ---* root@SERVER:/etc/dovecot/conf.d# doveadm auth test name.surname(a)subdomain2.domain2.cz Password: passdb: name.surname(a)subdomain2.domain1.cz*auth succeeded* extra fields: user=name.surname(a)subdomain2.domain1.cz root@SERVER:/etc/dovecot/conf.d# doveadm auth test name.surname(a)subdomain1.domain1.cz Password: passdb: name.surname(a)subdomain2.domain2.cz*auth failed* extra fields: user=name.surname(a)subdomain2.domain2.cz ----- I couldn't find any solution in the 2.4.1 documentation or on the Internet. AI didn't help. Thank you for your time and reply. Kind regards, Zbynek Grepl Hello List! I am testing auth-ldap.conf.ext on Dovecot 2.4.1 Debian 13 (trixie) Postfix 3.10.4 Dovecot 2.4.1-4 (7d8c0e5759) Everything works fine when I use only one ldap AD for authentication. I need to connect multiple ADs of our organizations which have separate ADs. The problem is that it always takes the last ldap_uris as working even if I put them in separate files. configuration - auth-ldap.conf.ext ### AD - ORG_ONE ldap_uris = [1]ldap://XX.YY.ZZ.CZ ldap_auth_dn = CN=ReaderAD,CN=Users,DC=domain1,DC=local ldap_auth_dn_password = secret_password_1 ldap_base = OU=ALFA,dc=domain1,dc=local ldap_version = 3 ### domena subdomain1.domain1.cz passdb ldap { ldap_filter = (&(objectClass=user)(mail=%{user|username}@%{user|domain})) ldap_bind = yes result_internalfail = return-fail } userdb ldap { driver = ldap ldap_filter = (&(objectClass=user)(mail=%{user|username}@%{user|domain})) fields { mail_path=/home/vmail/%{user|domain}/%{[2]ldap:mail}/Maildir mail_inbox_path=/home/vmail/%{user|domain}/%{[3]ldap:mail}/Maildir/ } } ### AD - ORG_SECOND ldap_uris = [4]ldap://AXX.AYY.AZZ.ACZ [5]ldap://BXX.BYY.BZZ.BCZ ldap_auth_dn = CN=ReaderAD,OU=ServiceAccount,OU=DELTA,OU=GAMA,OU=ALFA,OU=BETA,DC=domain2,DC=cz ldap_auth_dn_password = secret_password_2 ldap_base = OU=BETA,dc=domain2,dc=cz ldap_version = 3 ### domena subdomain2.domain1.cz passdb ldap { ldap_filter = (&(objectClass=user)(mail=%{user|username}@%{user|domain})) ldap_bind = yes result_internalfail = return-fail } userdb ldap { driver = ldap ldap_filter = (&(objectClass=user)(mail=%{user|username}@%{user|domain})) fields { mail_path=/home/vmail/%{user|domain}/%{[6]ldap:mail}/Maildir mail_inbox_path=/home/vmail/%{user|domain}/%{[7]ldap:mail}/Maildir/ } } --- doveadm auth test --- [8]root@SERVER:/etc/dovecot/conf.d# doveadm auth test [9]name.surname(a)subdomain2.domain2.cz Password: passdb: [10]name.surname(a)subdomain2.domain1.cz auth succeeded extra fields: [11]user=name.surname(a)subdomain2.domain1.cz [12]root@SERVER:/etc/dovecot/conf.d# doveadm auth test [13]name.surname(a)subdomain1.domain1.cz Password: passdb: [14]name.surname(a)subdomain2.domain2.cz auth failed extra fields: [15]user=name.surname(a)subdomain2.domain2.cz ----- I couldn't find any solution in the 2.4.1 documentation or on the Internet. AI didn't help. Thank you for your time and reply. Kind regards, Zbynek Grepl References Visible links 1. file:///tmp/tmpy9tgc9z4/ldap:/XX.YY.ZZ.CZ 2. file:///tmp/tmpy9tgc9z4/ldap:mail 3. file:///tmp/tmpy9tgc9z4/ldap:mail 4. file:///tmp/tmpy9tgc9z4/ldap:/AXX.AYY.AZZ.ACZ 5. file:///tmp/tmpy9tgc9z4/ldap:/BXX.BYY.BZZ.BCZ 6. file:///tmp/tmpy9tgc9z4/ldap:mail 7. file:///tmp/tmpy9tgc9z4/ldap:mail 8. mailto:root@server/etc/dovecot/conf.d# 9. mailto:name.surname@subdomain2.domain2.cz 10. mailto:name.surname@subdomain2.domain1.cz 11. mailto:user=name.surname@subdomain2.domain1.cz 12. mailto:root@server/etc/dovecot/conf.d# 13. mailto:name.surname@subdomain1.domain1.cz 14. mailto:name.surname@subdomain2.domain2.cz 15. mailto:user=name.surname@subdomain2.domain2.cz

1 0

doveadm search SENTSINCE and SENTBEFORE with timezone
by Philip Bondi 19 Oct '25

19 Oct '25

Hello to all: Would someone be able to comment on the correct method to specify timezone? [243] $ doveadm search -u user(a)domain.ca MAILBOX hover/INBOX-hover.mbx SENTSINCE "2024-01-01" SENTBEFORE "01-Jan-2025" fa6bc10562ee846855141f00d09efc50 4092 fa6bc10562ee846855141f00d09efc50 4093 fa6bc10562ee846855141f00d09efc50 4094 fa6bc10562ee846855141f00d09efc50 4095 fa6bc10562ee846855141f00d09efc50 4096 fa6bc10562ee846855141f00d09efc50 4097 [244] $ doveadm search -u user(a)domain.ca MAILBOX hover/INBOX-hover.mbx SENTSINCE "2024-01-01" SENTBEFORE "01-Jan-2025 00:00:00 -0400" Fatal: Invalid search date parameter: 01-Jan-2025 00:00:00 -0400

3 3

Missing imapc plugin
by scrivici＠winnbroker.it 17 Oct '25

17 Oct '25

Hi all, I have an installation (experimental, for now) of Dovecot CE 2.3.21 on Ubuntu and I'm trying to sync messages with an external IMAP server using doveadm sync, but I get the message "Plugin 'imapc' not found." It looks like the plugin isn't included in the version I'm using. Is that the case? Should I use a different version or download the plugin separately? Thank in advance for your help

2 7

Dovecot 2.4: dirsize removed - alternative for physical size quotas?
by jarland＠mxroute.com 16 Oct '25

16 Oct '25

We're upgrading our servers from Dovecot 2.3 to 2.4 and hit an issue with the dirsize driver removal. In 2.3, we switched from "maildir" to "dirsize" quota driver because maildir counts virtual (RFC822) sizes. This caused mailboxes to appear full before reaching actual disk limits - base64-encoded attachments make the virtual size larger than physical disk usage. For example, 8GB actual usage could show as 10GB+ in quota calculations. The dirsize driver solved this by calculating quota from actual file sizes on disk. In 2.4, dirsize is removed with the note "use quota-count instead." However, the count driver requires quota_vsizes=yes, which uses virtual sizes - the same problem we had with maildir. What's the intended solution in 2.4 for tracking quota based on physical disk usage rather than virtual message sizes? The available drivers seem to be: - count: requires virtual sizes - maildir: also uses virtual sizes via maildirsize - fs: requires OS-level per-user filesystem quotas Is there a way to configure count for physical sizes? Appreciate any guidance.

1 0

Feature request: allow Settings variables to be used in ldap_auth_dn
by Kristian 16 Oct '25

16 Oct '25

Hi, with the migration to 2.4 my setup for LDAP authentication stopped working. I used to have different LDAP config files for every virtual domain. Each domain has an domain-specific service user with read access, that is used by dovecot. There is no service user with access across domains. With the current 2.4 configuration I only see the possibility to configure one user in the variables ldap_auth_dn ldap_auth_dn_password In ldap_base I am able to use settings variables ldap_base = ou=people,dc=%{user | domain | ldap_dn} which allows me to select the right LDAP tree depending on the domain of the user. I would need this also in ldap_auth_dn to keep my current setup. I can see, that this is specific to my setup without an service user with access to all domains, and that this ignores the problem of having different passwords for different domain users, but it would at least keep some compatibility to working 2.3 setups. Kind regards Kristian

1 0

auth_cache_ttl and sieve_enabled
by Arkadiusz Miśkiewicz 14 Oct '25

14 Oct '25

Hi, auth_cache_ttl docs https://doc.dovecot.org/main/core/config/auth/caching.html say: " # We also try to handle password changes automatically: If user's previous # authentication was successful, but this one wasn't, the cache isn't used. " but seems that this doesn't work in such way for sieve_enabled flag changes. So changing sieve_enabled from 0 to 1 means user still has to wait until cache expires and new value is noticed by dovecot... which is not desired. Could sieve_enabled be also treated in similar way as incorrect authentication? -- Arkadiusz Miśkiewicz, arekm / ( maven.pl | pld-linux.org )

1 0

Ssl_cert keyword ?
by MellorCH 10 Oct '25

10 Oct '25

> V2.4.1 what’s the latest syntax for this please? Ssl_cert V2.4.1 what's the latest syntax for this please? Ssl_cert

3 2

SSL initialization error - Dovecot 2.4.1, Fedora 43 beta
by Jason J.G. White 09 Oct '25

09 Oct '25

I'm trying Dovecot 2.4.1 under Fedora 43 beta with all packages up to date. I've encountered several issues after attempting to update my configuration. The main problem is this: whenever a client attempts to connect, I get the following error Error: Failed to initialize SSL connection: Couldn't initialize SSL server context: Can't load SSL certificate (ssl_server_cert_file setting): error:0A00018F:SSL routines::ee key too small: I'm using a SECP384R1 key issued by letsencrypt.org, which should satisfy security checks. My Web searches have not identified a relevant bug report. Have others encountered this issue? If it's distribution-specific, I can visit the relevant bug tracker. Thank you for your consideration.

5 8